Three weeks ago I ran a diligence pipeline over a token that had raised real money this cycle. Not a meme coin โ a project with a forty-page documentation site, a foundation entity in a friendly jurisdiction, two tier-two exchange listings, and a narrative about "institutional-grade infrastructure" that had been repeated so often it had begun to sound like a fact rather than a claim. I fed everything I had into the structured frame I've used since 2019: technical surface, token economics, market structure, ecosystem position, regulatory posture, team and governance, risk surface, narrative-and-expectation gap, supply-chain transmission. Nine dimensions. The same nine I run on everything.
The output came back with every field marked N/A.
I want to be honest about my first reaction, because the reaction is the story. I assumed the tool was broken. I went hunting for a bug โ a parsing failure, a schema mismatch, something dumb and mechanical. What I found instead was that the pipeline had done precisely what I built it to do. It refused to populate fields that had no source behind them. So I went back to the project's own documentation to prove the tool wrong, and I couldn't. Every heading was there. Every claim had a heading. Nothing was underneath.
That was the finding. Not a bug. A mirror.
I have been doing this long enough to recognize the shape. In 2017, at twenty, I spent six months manually reading the genesis block code of five ICO projects โ Tezos and MakerDAO among them โ and wrote a forty-page thesis arguing that smart contracts would relocate trust from institutions to mathematics. I was wrong in an interesting way, which is the best kind of wrong. What I actually learned in those six months wasn't about cryptography. It was that the whitepapers I was reading had a consistent architecture: an elegant philosophical preamble, a technical section with real diagrams, and a token distribution table at the back. The preamble was beautiful. The diagrams were real. The distribution table was where the actual project lived, and it was almost always the least explained page in the document.
That was my first encounter with a field that exists but carries nothing.
Six years later, in the 2020 DeFi summer, I put fifteen thousand Australian dollars of my own savings into an unaudited yield farm and lost all of it in forty-eight hours. I spent the next three months reverse-engineering the exploit and publishing every step in a public repository. That post-mortem taught me something the whitepapers never could: on a blockchain, I could trace every single wei from my wallet to the attacker's contract. Perfect provenance. Absolute, cryptographic, timestamped, undeniable provenance.
And it still didn't tell me why. Provenance is not understanding. A chain can prove that a thing happened with more rigor than any institution in human history, and remain completely silent on what the thing meant.
So when the pipeline returned nine N/A's, I didn't experience it as a failure of research. I experienced it as the most honest document I had read in a bull market in eighteen months.
Here's the thing about the N/A report that I keep circling back to: there are two distinct ways a field can carry zero information, and crypto in 2026 is drowning in both.
The first is the empty field. That's what I got. A heading with nothing beneath it. It's ugly, it's unsatisfying, and it is at least honest, because an empty field is visibly empty. You can't mistake it for content. The second is far more dangerous โ the field that is always filled, always filled with the same thing, and therefore conveys nothing at all. In information terms, a field whose value never changes carries zero bits regardless of how many words it contains. A slide that says "progressive decentralization roadmap" and has said exactly that for twenty-six consecutive months is not a data point. It's a decoration that has learned to sit in the shape of a data point.
The empty field and the boilerplate field are the same failure wearing different clothes. One admits it, one performs confidence. Markets reward the second one, which is why we get so much of it.
Let me show you what a live field actually looks like, because this is the part that gets skipped. A field carries information when it satisfies three conditions: it has a date, it has a source, and it has a falsifier โ some observable outcome that would prove it wrong. Take the single most consequential field in any governance analysis: upgrade authority. Who can change the rules after the token is live?
Almost every project will answer this question somewhere. Very few answer it in a form that meets the three conditions. "Governance controls the protocol" fails all three. "The protocol is upgradeable by a 4-of-7 multisig held by the founding team, with signers disclosed here, and this arrangement is scheduled to change at a date that has slipped twice" โ that passes. It's dated, it's sourced, and it's falsifiable, because you can go look at the multisig on-chain and see whether the signer set matches the claim. And that's the whole game. In my experience auditing governance structures across roughly forty protocols, the upgrade authority field is the one that most often reads as filled while actually being blank, and it is also the field that determines whether every other field in the report means anything.
This is where I part company with the version of myself who wrote that thesis at twenty. Code is law doesn't hold, and it never did โ not because the code fails, but because the right to change the code always sits somewhere, and that somewhere is almost always a handful of keys held by a handful of people who know each other. You can have the most elegant on-chain voting apparatus in the world and it doesn't matter, because the multisig can upgrade the voting contract. The governance forum is where the argument happens. The multisig is where the decision happens. Any analysis that documents the first and leaves the second as N/A has documented the theatre and skipped the building.
Now widen the lens, because this problem is not confined to governance templates. It's structural, and it lives in the infrastructure layer.
When Celestia's whitepaper landed in my reading pile during the 2022 collapse โ I had just laid off my only employee and was sitting in a Sydney flat with a lot of time and very little revenue โ the idea that pulled me in wasn't scalability. It was a distinction I had never seen stated so cleanly: data availability is not validity. A data availability layer can guarantee that the data behind a block was published and is retrievable by anyone who wants it. It says nothing, absolutely nothing, about whether the transactions inside that block are correct. Those are two different guarantees, produced by two different mechanisms, answering two different questions.
That distinction is the entire architecture of the modular thesis, and it is also a perfect miniature of the problem I've been describing. Availability is the field. Validity is the field. They look adjacent on a diagram. They are not the same claim.
Ethereum's blob upgrade made the same distinction concrete for a much larger audience. After blobs arrived, rollups could post data to the base layer cheaply, and fees collapsed across the L2 ecosystem โ a genuine engineering achievement. But notice what actually changed. The cost of publishing data fell by orders of magnitude. The cost of publishing meaningful data did not change at all. We made it dramatically cheaper to put things on-chain, and we did not make it one bit easier to know what those things are worth. Cheap availability is not the same as cheap truth, and every cycle we confuse the two because the first one produces a chart that goes up.
I keep thinking about the library analogy, because it's the one that survives contact with non-technical readers. A data availability layer is a library that can prove, mathematically, that every book you ordered is physically on the shelf, in the correct slot, retrievable within a guaranteed time. That's a real and hard guarantee and it's genuinely valuable. What it cannot tell you is whether any of the books are true. Availability is the librarian confirming the shelf. Truth is a different building.
So what do we actually have? We have the most powerful provenance machine ever built, and we have almost no infrastructure for meaning. My 2020 post-mortem is the clearest example I own. I could reconstruct the exploit transaction by transaction. I could tell you the exact block, the exact function selector, the exact reentrancy path, the exact moment my fifteen thousand dollars left my control. Every fact in that document is verifiable by anyone with an archive node. And none of those facts explain the thing I actually needed to understand, which is that I had sized a position at one hundred percent of my savings because I was excited, and excitement is not a risk model. The chain recorded the consequence perfectly. It had nothing to say about the cause, because the cause happened in my head.
That's the asymmetry. Blockchains are extraordinarily good at recording what happened and structurally incapable of recording why it mattered. Every piece of research that pretends otherwise is filling a field it cannot source.
Which brings me to the incentive question, and this is where the bull market does its damage. Research in a rising market is not produced to be true. It is produced to be consumed. Those are different objectives with different failure modes. If your product is coverage, then the worst possible outcome isn't being wrong โ it's being absent. An analyst who publishes nothing about a hot token loses mindshare. An analyst who publishes nine N/A's loses mindshare and looks lazy. An analyst who publishes a confident nine-dimension report with three fields quietly invented and six filled from the project's own marketing copy gets quoted, gets invited to the Spaces, gets the retainer.
I've watched this generate what I think of as coverage debt. Every cycle, the industry accumulates a stock of documents that look like analysis and function as marketing, and the debt compounds because the next analyst cites the previous report rather than the primary source. Two or three citation hops later, a claim that originated in a project's own deck is being treated as an independently verified fact. The chain of custody is broken and nobody notices, because the document is well-formatted and the font is nice.
Formatting is the most effective camouflage in this industry. A complete-looking table over an incomplete dataset is more dangerous than no table at all, because the completeness of the form gets mistaken for the completeness of the content.
I want to be specific about where I see the boilerplate field doing the most damage right now, because vague criticism is its own kind of empty field.
Start with the sequencer. Every serious rollup has a decentralization roadmap. I have been reading these documents since 2023 and I can tell you that the language has barely moved. "Sequencer decentralization" has been a slide for two years, and in that time the actual architecture of most major rollups has not changed in the way the slide implies: a single operator orders transactions, that operator is a company, and the escape hatch โ forced inclusion through the base layer โ exists and is real but is a fallback, not a substitute for ordering rights. The honest version of the field would read: "one node orders your transactions today; you can force a transaction through L1 if it censors you, at the cost of latency and fees; here is the date we expect this to change and here is what would have to be true for that date to be credible." That's a filled field. What we get instead is a roadmap.
The same pattern shows up in payments, and it bothers me more because the stakes are human. I have spent a lot of time on stablecoin corridors over the last two years, and the research on emerging-market adoption is consistently full of prose about financial inclusion and ideology and empty on the only variable that explains the behaviour: local currency inflation and the practical availability of dollar-denominated savings. Nobody in Buenos Aires or Lagos or Istanbul is using a dollar stablecoin because they read a decentralization manifesto. They're using it because the alternative is watching their savings dissolve in real time. The driver is monetary, not philosophical. Yet if you read the average institutional report on payment stablecoins, you'll find a well-developed section on the ideology of decentralization and a blank where the inflation data should be. The field that matters is empty and the field that flatters is full.
So the question I actually want to answer is not "why is crypto research bad." It's the engineering question: can we make N/A a first-class object?
I think we can, and I think the primitives already exist. Attestation registries are the obvious candidate โ a schema where a claim is a structured record with a subject, an issuer, a timestamp, and a reference to evidence. The important property isn't that attestations are on-chain, because on-chain adds nothing to truth. The important property is that an attestation makes the absence of a claim visible. If every protocol in a category were expected to publish a standard governance attestation, then a missing attestation becomes a signal, and a stale attestation becomes a signal, and the difference between "we haven't published" and "we published and it says something you won't like" becomes legible to anyone with a block explorer.
The thing I find genuinely beautiful about this is that it inverts the incentive. Right now, the project that says nothing and the project that says something misleading are indistinguishable to a casual reader, because both produce a documentation site with headings. If absence were visible by default, silence would become expensive. The goal isn't to put truth on-chain โ truth in blockchain isn't a property the chain can grant. The goal is to make the shape of our ignorance visible, which is a much more achievable engineering problem.
And there's a version of this that costs almost nothing. Date every claim. Cite every source. Name the falsifier. Three fields, no new protocol, no token, no governance vote. I have started refusing to work with projects that won't do it, which is a luxury of having thirteen years of context and a reputation that survives saying no. Not everyone has that luxury, which is precisely why the norm has to come from the reader side rather than the writer side. Readers have to start asking the boring question: what would prove this wrong, and when was it last checked?
I want to make an argument against myself here, because the N/A report has a shadow and I'd rather name it than have someone else name it for me.
Refusing to analyze without a complete information base is honest. It is also, sometimes, a hiding place. If the standard for speaking is a fully sourced nine-dimension analysis, then nobody ever speaks, and the field is left to the people who don't care about sourcing. That's not integrity โ that's abdication with good manners. I have watched extremely careful analysts produce immaculate N/A's for years while louder, sloppier people set the narrative, and the careful ones console themselves that they were right to wait.
The truth is that judgment under uncertainty is the job. Nobody in this industry has ever had a complete information base, including the people who were right. My 2020 self didn't fail because he lacked data โ the audit was one click away and I didn't read it. He failed because he didn't size the position. The lesson wasn't "gather more information." The lesson was "make a call with the information you have and state how much you'd bet on it."
So the report I actually want isn't nine N/A's. It's nine fields where three are sourced, four are explicitly marked as inference with a confidence level attached, and two are marked unknown with a note on what would resolve them. That document is honest and it's also useful, and the difference between it and the empty table is the difference between a researcher and a filing cabinet. Calibrated assertion is a skill. Blanket refusal is a posture. The industry has far too much of the second and nowhere near enough of the first.
The other thing I'd push back on is the fantasy that the empty report revealed a broken pipeline. It didn't. It revealed a project whose entire public surface was a template. That's a much more serious finding, and it's one that no amount of tooling can produce for you โ somebody has to actually read the forty pages and notice that the distribution table is missing.
Which is where I've landed after three weeks of turning this over. The most valuable document I received this quarter contained no information at all, and it was more useful than the nine well-formatted reports I read before it, because it told me exactly where the boundary of my knowledge was and refused to blur it.
We didn't build this industry to produce prettier tables. We built it to make claims checkable, and checkability has a prerequisite that no amount of infrastructure can supply: someone has to be willing to write down that they don't know.
So here's the question I'm carrying into the rest of this cycle, and I'd like you to sit with it rather than answer it quickly. When the next project hands you a complete-looking document with every heading in place, will you count the headings โ or will you go looking for the one field that's silently empty, the one where the multisig sits, the one nobody dated, the one that would tell you who can change the rules after you've bought in?
Because that field is always there. It's just never labelled N/A.
